Introduction

Synaptic plasticity refers to a wide range of synaptic changes occurring in neurons in response to specific physiological events, with the final goal of adaptation to new information. Plasticity represents the set of specific adaptations occurring at synapses and junctions allowing the communication between neurons. The original idea that neurons might communicate each other, and change their synaptic activity in response to an event occurring in a different neuron was proposed in the 1949 by the Canadian psychologist Donald Hebb. Hebb in the “Organization of Behavior. Wiley: New York; 1949”, for the first time, theorized the basis for the concept of Synaptic Plasticity (Hebb 1949); “When an axon of Cell A is near enough to excite a Cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A’s efficiency, as one of the cells firing B, is increased”.

In 1973, this theory was confirmed by the first report demonstrating the possibility to induce Long-Term Potentiation (LTP) of synaptic transmission in the dentate area of anaesthetized rabbits under perforant path stimulation (Bliss and Lomo 1973). Bliss and colleagues demonstrated both in anaesthetized (Bliss and Lomo 1973) and unanesthetized rabbits (Bliss and Gardner-Medwin 1973) that two different synapses can communicate with different strength and this phenomenon is not static, but rather can change in both short- and long-term manners. The complex range of events that can be included into the definition of Synaptic Plasticity comprises synaptic signal transmission (i.e., Short- and Long-Term Plasticity) as well as structural alterations (i.e., morphological spine density change). These neuronal changes can be the consequence of the physiological maturation during development, the acquisition of new information during learning, or can be induced by brain damage.

Synaptic plasticity changes in Parkinson’s disease

In neurodegenerative disorders, such as Parkinson’s disease (PD), altered synaptic plasticity is the consequence of aberrant cellular and molecular cascades leading to neuronal death, and it is at the same time the cause of behavioral and cognitive dysfunctions. In PD, as well as in other neurodegenerative disorders, synaptic dysfunctions precede the neuronal loss of decades; therefore, their early identification might help to identify the pathological cascade leading to neuronal death and the neuronal basis of early behavioral symptoms, such as non-motor symptoms (Ayala et al. 2017; De Leonibus et al. 2007).

This privileged position makes aberrant synaptic plasticity the target of both disease modifiers and symptomatic rescue strategies, and of their possible side effects. There is no cure for PD, but only symptomatic treatments that help the patients to manage the motor symptomatology. Among the different therapeutic strategies, dopamine (DA) replacement therapy with levodopa (l-DOPA) is still the only pharmacological treatment that has been proved to improve the motor symptoms (Olanow and Schapira 2013). However, chronic treatment with l-DOPA has also major side effects, among which dyskinesia, the occurrence of involuntary movements, is the most invalidating. Therefore, understanding the synaptic mechanisms through which the therapeutic effects of l-DOPA are converted into side effects is crucial to prevent them and to design novel drugs void of dyskinesia effects.

l-DOPA has also other important non-motor side effects, such as impaired cognition and increased risk of psychotic-like episodes (Bastide et al. 2015; Voon et al. 2009, 2017). A range of impulse control disorders including gambling, compulsive shopping, compulsive sexual behaviors, and binge eating, occurs in about 17% of PD patients on dopaminergic medications. In addition, the compulsive use of medication L-DOPA and DA agonists is also associated with an exacerbation of L-DOPA-induced dyskinesias (LIDs) (Voon et al. 2017). Some of these pathological side effects of chronic treatment with l-DOPA might be due to aberrant expression of alpha-synuclein in PD patients. Alpha-synuclein protein accumulation and misfolding in Lewy Body (Spillantini et al. 1997), is the main histological hallmark of a series of late-onset neurodegenerative disorders, including PD, other parkinsonisms such as dementia with Lewy bodies, and multiple system atrophy and other rare diseases (Barker and Williams-Gray 2016). These pathologies are generally referred as synucleinopathies. The strict association between aberrant neuronal alpha-synuclein accumulation and PD is demonstrated by the fact that patients receiving a first diagnosis of dementia with Lewy Body (DLB) have 70% possibility to develop motor PD; viceversa patients with first diagnosis of PD have higher risk of developing dementia if they also have altered beta-amyloid peptides levels in the cerebro spinal fluid (CSF) (Parnetti et al. 2014). The use of l-DOPA to control motor symptoms in DLB patients with PD or in PD patients with dementia is extremely limited as it worsens cognitive symptoms and hallucinations, likely due to its action on cortical regions (Poewe 2005); understanding the aberrant synaptic mechanisms induced by alpha-synuclein and the effects of l-DOPA not only in the striatum, but also in the other major components of the corticolimbic system is fundamental to design personalized therapeutic approaches that takes into account the complex symptomatology associated to PD.

Here, we show the recent scientific contributions in the field of synaptic plasticity alteration in PD and LIDs with particular attention to the structural events underlying these disorders and the related synaptic basal ganglia (BG) network alterations in pre-clinical and clinical studies. Most of the reviewed experimental literature is based on pharmacological animal models of PD, in particular with the use of 6-hydroxydopamine (6-OHDA) and of 1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP).

Altered synaptic plasticity in experimental LIDs and dyskinetic PD patients

The two main forms of synaptic plasticity expressed in the brain are LTP and Long-Term Depression (LTD) (Malenka and Bear 2004; Calabresi et al. 1992b, c; Lovinger et al. 1993).

Synaptic plasticity alterations, either in experimental animal models or in PD patients, have been widely investigated. It has been consistently reported that the lack of dopaminergic tone in experimental rats lesioned with 6-OHDA causes the loss of synaptic plasticity, i.e., LTD and LTP, in striatal projection neurons (SPNs) (Calabresi et al. 1992a; Centonze et al. 1999; Belujon et al. 2010; Cerovic et al. 2015; Shen et al. 2015). Different stimulation protocols can be used to induce different forms of synaptic plasticity within the striatum in physiological and pathological conditions. In particular, high-frequency stimulation (HFS) protocol consists in three trains at 100 Hz for 3 s, 20 s interval, while spike-timing-dependent plasticity (STDP) is a pairing stimulation in which a synapse is activated by stimulating a pre-synaptic neuron shortly before or after making the post-synaptic neuron fire by injection of a short current pulse. These two different approaches represent the most useful tools to unravel alterations of synaptic and molecular pathways, underlying parkinsonian and dyskinesia conditions (Calabresi et al. 1992a; Centonze et al. 1999; Belujon et al. 2010; Cerovic et al. 2015; Shen et al. 2015; Augustin et al. 2014; Fino et al. 2010; Shen et al. 2008; Surmeier et al. 2007).

The lack of synaptic plasticity in DA denervated animals has been confirmed also in PD patients. The Paired Associative Stimulation (PAS) approach is a useful method to study physiological and aberrant plasticity in the human M1 motor cortex (Stefan et al. 2000, 2004; Wolters et al. 2003). In human studies, the measurement of Motor Evoked Potential (MEP) amplitude is considered a reliable analog of LTP (Stefan et al. 2000). This approach has been used to demonstrate the loss of LTP in the motor cortex of PD patients in OFF medication, compared with the potentiation of PAS parameters measured in control subjects (Morgante et al. 2006; Ueki et al. 2006). The synaptic plasticity in the cortex of PD patients is modulated by dopaminergic tone (Morgante et al. 2006; Ueki et al. 2006; Molina-Luna et al. 2009). In PD patients in ON medication state, chronic l-DOPA treatment restores, in the first phase of the disease, the dopaminergic levels and normalizes cortical M1 LTP (Morgante et al. 2006; Kawashima et al. 2013).

Although, the previous experimental and human studies (Morgante et al. 2006; Ueki et al. 2006; Molina-Luna et al. 2009; Kawashima et al. 2013) have identified a role of DA in synaptic plasticity in the BG input structures and motor cortex, their role in Substantia Nigra pars reticolata (SNr) has been little investigated. Prescott and colleagues extended these observations in a different BG structure of PD patients by the use of Deep Brain Stimulation (DBS) approach (Prescott et al. 2009). In this context, Prescott and Hutchison point out the attention to the possibility to record synaptic plasticity in SNr using extracellular field potentials (fEPs) in PD patients in OFF and ON medication. The application of HFS protocol in the SNr of parkinsonian patients in OFF medication state fails to induce LTP of fEPs amplitude, while in the ON state, this potentiation is present (Prescott et al. 2009).

The dysregulated control of BG circuitry and the lack of physiological synaptic plasticity within the corticostriatal terminals ultimately cause a wide range of molecular and synaptic alterations underlying the movement features of PD and l-DOPA-induced (Albin et al. 1989; Calabresi et al. 2016; Wang and Zhang 2016).

The treatment with l-DOPA represents the gold-standard symptomatic PD therapy (Mercuri and Bernardi 2005; Olanow and Schapira 2013). l-DOPA succeeds to counteract the effect of dopaminergic decrease allowing amelioration of motor symptoms during the initial phases of the disease, when not all the dopaminergic terminals are lost (Calabresi et al. 2015). Unfortunately, the so-called “honey moon period” of l-DOPA therapy deteriorates rapidly, after a variable period of time for each patient, and progressively develops in motor fluctuations (Olanow et al. 2004; Calabresi et al. 2015).

The development of LIDs has been associated to impaired bidirectional synaptic plasticity at corticostriatal synapses of SPNs in 6-OHDA parkinsonian rodents (Picconi et al. 2003; Picconi et al. 2008; Cerovic et al. 2015; Belujon et al. 2010). The term “bidirectional plasticity” refers to the possibility to express flexible synaptic plasticity; i.e., the possibility to induce, with the same HFS protocol, a depression or a potentiation of synaptic transmission, depending on the different receptor state or to bring back (“down-scale”) to control levels the potentiated synapse after a HFS protocol. In particular, these pre-clinical studies paved the way to the investigation of the physiological phenomenon of depotentiation of LTP in the striatum, which is lost in hyperkinetic conditions such as LIDs (Picconi et al. 2003; Calabresi et al. 2016; Prescott et al. 2014; Martella et al. 2009; Centonze et al. 2006). Depotentiation is thought be a physiological mechanism to erase unessential information in a previously potentiated synapse and bring it back to basal level to be ready to respond to a next input (Picconi et al. 2003; Calabresi et al. 2016). The induction of dyskinetic movements and the loss of depotentiation are actually two sides of the same coin; parkinsonian condition erases the capability to induce LTP (Centonze et al. 1999; Ueki et al. 2006), while the treatment with l-DOPA either in parkinsonian animals or in PD patients restores LTP induction both in non-dyskinetic and in dyskinetic subjects (Picconi et al. 2003; Picconi et al. 2008; Prescott et al. 2009). However, the LTP in dyskinetic subjects, differently from that of non-dyskinetic, does not depotentiate under the application of a low frequency stimulation (LFS) protocol. In other words, l-DOPA restores synaptic plasticity in denervated animals, but in dyskinetic animals, this form of synaptic plasticity is not “plastic”, as it is not sensitive to homeostatic synaptic inputs. Both pre-clinical and clinical studies demonstrate that the physiological phenomenon of bidirectional or homeostatic plasticity indirectly correlated to LIDs development (Picconi et al. 2003; Picconi et al. 2008; Prescott et al. 2014; Huang et al. 2011). In addition, it has been shown that parkinsonian patients in OFF medication state, independently of their LIDs state, do not express potentiation of MEP amplitude measured with PAS method. Interestingly, l-DOPA treatment restores normal LTP in non-dyskinetic patients, but fails to restore physiological plasticity in dyskinetic subjects (Morgante et al. 2006).

The absence of downscaling mechanism in the striatal SPNs synapse, in pathological conditions such as PD and LIDs, has been associated to dopaminergic D1 receptor downstream pathway hyperactivation. In particular, dyskinetic condition is associated to an abnormal regulation of D1/protein kinase A (PKA)/dopamine- and cAMP-regulated neuronal phosphoprotein 32 kDa (DARPP-32) and Ras–extracellular-signal-regulated kinase (Ras–ERK) signaling pathways (Bido et al. 2015; Cerovic et al. 2015; Fasano et al. 2010; Picconi et al. 2003). An additional confirmation of the relationship between the loss of depotentiation and abnormal DA-related D1 downstream activation has been demonstrated in animals chronically treated with cocaine (Centonze et al. 2006).

Recently, the use of STDP protocol allows to analyze, from a different point of view, the specific role of direct and indirect striatal pathways in experimental parkinsonian and dyskinetic mice, carrying on eGFP targeting D1- and D2-expressing SPNs (Thiele et al. 2014). This study confirmed that bidirectional plasticity (LTD, LTP, and depotentiation) is present in both direct and indirect striatal neurons. The parkinsonian mice lose this form of homeostatic plasticity; i.e. SPNs exhibit only unidirectional plasticity, regardless of the stimulation paradigm (LTD or LTP). After chronic l-DOPA treatment, the mice that do not develop dyskinesia show normal LTP and depotentiation both in D1- and D2-expressing SPNs. Conversely, dyskinetic mice present a restoration of LTD in the D2-expressing neurons (indirect pathway) and of LTP in the D1-expressing neurons (direct pathway) (Thiele et al. 2014).

Other mechanisms might be responsible of LIDs, as demonstrated by recent studies supporting a role of muscarinic M4 receptors in preventing LIDs; in this study the activation of M4R signaling, by a positive allosteric modulator, in parkinsonian mice chronically treated with l-DOPA interferes with the development of dyskinetic movements and blocks bidirectional plasticity in the direct-pathway neurons (Shen et al. 2015).

Although these seminal studies provided the first neuronal and molecular basis underlying LIDs, more recent experimental evidence is adding complexity to the neuroanatomical and neurophysiological micro-circuitry in the BG involved in PD and LIDs (Cerovic et al. 2015; Belujon et al. 2010; Calabresi et al. 2016; Morgante et al. 2006).

Belujon and colleagues (Belujon et al. 2010) suggested a different pattern of plastic changes in parkinsonian rats chronically treated with l-DOPA. In particular, this study measured corticostriatal LTD in vivo through extracellular recordings of striatonigral “direct” pathway versus striatopallidal “indirect” pathway neurons; HFS protocol induces normal LTD in both striatal neuronal pathways of control animals (Belujon et al. 2010). Interestingly, the application of a depotentiation protocol induces a reverse of LTD only in striatonigral neurons, while neuronal activity of striatopallidal cells appears to be less prone to come back to control levels (Belujon et al. 2010). The striatonigral neurons recorded in parkinsonian rats express a normal LTD but the possibility to reverse its expression is lost. Most importantly, long-lasting depression of striatopallidal neurons is fully abolished in parkinsonian condition, accounting for their possible higher vulnerability. Notably, chronic treatment with l-DOPA, in rats not presenting LIDs, restores depotentiation after LFS in striatonigral neurons and induces a shift from LTD to LTP in striatopallidal neurons; the latter is not sensitive to the LFS downscaling protocol. This pattern of activity differs from that of dyskinetic animals exclusively for the effects of LFS downscaling protocol; in animals with LIDs the LFS protocol does not reverse the LTD in striatonigral neurons, but it reverse the LTP in striatopallidal neurons (Belujon et al. 2010).

All together, these in vivo and in vitro data suggest that l-DOPA favors hyperexcitability of the indirect pathway, which if contrasted by flexible downscaled processes in the direct pathway might act with therapeutic result; in contrast, if it is associated to the lack of bidirectional plasticity in the direct pathway, it gives rise to LIDs. Thus, LIDs might be the results of the lack homeostatic mechanisms in the D1 pathway, favored by a hyperactivation of the indirect pathway.

Structural synaptic plasticity in PD and LIDs

Dendritic spines represent the hardware structure on which synaptic plasticity processes run; therefore, they set the limit and the potentiality of the network. The different aspects of spine morphology differently affect synaptic transmission and plasticity, depending on the specific organization of the network.

In the striatum, most if not all, cortical glutamatergic inputs terminate on the head of dendritic spines of SPNs (Deutch et al. 2007; Raju et al. 2006, 2008). Thalamostriatal inputs also reach the same part of SPNs, but they target different types of striatal spines; for instance, the volume as well as the post-synaptic density (PSD) zone of the striatal spines, receiving cortical inputs, are larger as compared to those innervated by thalamic inputs, and they predominantly express the Vesicular Glutamate Transporter 1 (VGluT1), rather than VGluT2. Dopaminergic inputs, as well as cholinergic interneurons contacts, predominantly reach the neck of the spine consistent with a modulator role on the effects of glutamatergic inputs.

Almost complete dopaminergic denervation (> 90% loss) produces long-lasting changes in the morphology of SPNs neurons; this finding has been consistently reported across different species (monkeys, rats, and mice) and by the use of different kinds of toxins (reserpine, 6-OHDA and MPTP). Experimental parkinsonisms lead to reduced spine density, which ranges around 20–50% (Day et al. 2006; Gagnon et al. 2017; Ingham et al. 1993; Shen et al. 2007; Suarez et al. 2014; Toy et al. 2014; Villalba et al. 2009), according to the findings shown in the human putamen of PD patients (McNeill et al. 1988). Partial DA depletion has been shown to lead to reduced spine density in a progressive model of PD in monkeys and mice treated with MPTP (Villalba et al. 2009), but not in partially spared striatal area in an acute model of PD (6-OHDA lesion) (Day et al. 2006). These changes in partial DA denervated model might underlay early behavioral dysfunctions in PD, and need further investigation.

Reduced spine density has been identified in both types of SPNs expressing D1 or D2 DA receptors in the vast majority of studies (Day et al. 2006; Gagnon et al. 2017; Suarez et al. 2014; Toy et al. 2014; Villalba et al. 2009). Interestingly, DA denervation also reduces spine density in SPNs expressing D1–D2 heterodimers, which represent the 2% of SPNs in the dorsal striatum (Gagnon et al. 2017).

These findings suggest that DA inputs to dendritic spines of SPNs exert a necessary neurotrophic action for their maintenance. Interestingly, a parallelism in the proportion between the number of spines receiving DA inputs and those that are lost after DA denervation supports a possible neurotrophic action of DA on spines growth (Ingham et al. 1998). However, as previously suggested, the neurotrophic action of DA input on dendritic spines might be due to DA containing cells, which also contain other neurotrophic factors such as brain derived neurotrophic factor (BDNF), which is crucial for spine growth and maintenance (Seroogy et al. 1994). Other mechanisms have also been suggested to participate in this process, mainly related to glutamate-dependent excitotoxicity. Although most of the lost spines also contain asymmetric excitatory contacts, ultrastructural changes indicative of increased strength of glutamatergic transmission have been identified. For instance increased spine volume and PSD perforation have been reported after DA denervation in the remaining SPNs spines, which positively correlate with increased neuronal excitability (Villalba et al. 2009).

Interestingly, experimental evidence in pups (17–25 days) suggested a selective pruning of dendritic spines in the indirect pathway, which is mediated by aberrant activity of Cav1.3 Ca2 + channels (Day et al. 2006). Whether this same mechanism applies also to adult animals and DA denervated models showing spine pruning in D1-expressing SPNs, remains to be clarified.

Regardless of the lack of clear evidence on the molecular mechanism leading to synaptic pruning induced by DA denervation, some experimental findings suggest that DA replacement therapy with chronic l-DOPA treatment, as well as other DA agonists, can increase spine density not only in the striatum but also in the primary motor cortex and in the ventral striatum (Funamizu et al. 2017; Suarez et al. 2016; Ueno et al. 2017). These findings are quite surprising considering that they have been obtained also in animals showing dyskinetic movements with almost complete depletion of striatal DA content (Suarez et al. 2014); whether these new spines are fully functional has never been addressed.

In those studies where changes in synaptic density have not been associated to LIDs, consistent alterations in spines morphology have been identified, in particular spine head enlargement, related to increased debrin immunoreactivity (Funamizu et al. 2017; Nishijima et al. 2017; Ueno et al. 2017; Nishijima et al. 2013). Debrin is a F-actin binding protein, almost exclusively expressed in neurons, commonly used as a marker of excitatory input (Nishijima et al. 2013). A still controversial issue is whether these changes in glutamatergic synapses are cellular subtype or afferent subtype specific. Dyskinetic abnormalities seem to be associated to selective abnormal increase in corticostriatal contacts, but not in the thalamostriatal inputs (Zhang et al. 2013), on the indirect pathway (Suarez et al. 2016; Schuster et al. 2009).

Synaptic and structural plasticity in experimental LIDs and dyskinetic PD patients

Few studies combined morphological changes with electrophysiological alteration in synaptic activity, which do not allow to establish a direct causal link between the two types of plastic events. Increased head spine and spine density positively correlate with increased amplitude and frequency of miniature excitatory post-synaptic currents (mEPSCs), respectively (Segal 2010). Furthermore, LTP and LTD protocols lead to increased and decreased spine head size, respectively (Matsuzaki et al. 2004; Yasumatsu et al. 2008). Therefore, based on the morphological data, it might be inferred that 6-OHDA lesion should lead to reduced frequency and increased amplitude to excitatory transmission in SPNs neurons. Notably, the lack of a correct dopaminergic tone reflects, in 6-OHDA animals, an increase of striatal glutamatergic transmission (Centonze et al. 2005; Cepeda et al. 2001; Gubellini et al. 2002; Tozzi et al. 2011; Tang et al. 2001; Maccarrone et al. 2003) (Fig. 1).

Fig. 1
figure 1

Structural and synaptic changes in striatal projection neurons. In physiological conditions (a) motor information processing causes a trophic effect on striatal spines, an increase in glutamatergic pre-synaptic vesicles, and a normal synaptic plasticity (b). In Parkinson’s disease (PD), the loss of LTP is accompanied by a significant spine pruning and alterations in glutamatergic receptors expression and glutamate release (c). Chronic levodopa treatment restores LTP by replacing DA release, but do not re-establish depotentiation as a consequence of pre-synaptic mechanisms of DA release regulation and post-synaptic increased spine head size associated to increased glutamatergic activity (d)

Full size image

l-DOPA treatment leads to initial restoration of synaptic plasticity and behavior likely through some degree of spared mechanism governing synaptic phasic DA release, such as impulse-dependent DA release and proper re-uptake through the dopamine active transported (DAT). However, it could be hypothesized that chronic treatment leads to a shift toward tonic DA release; tonic DA release, under chronic l-DOPA treatment, might induce a shift of DA action on the hyperactive D2 pathway as evidenced in in vivo studies in dyskinetic rats (Belujon et al. 2010) and as evidenced by the shift from LTD to LTP and by increased spine density (Suarez et al. 2016). At the same time, it might lead to a lack of phasic control of DA D1 receptors activation, which is necessary for homeostatic bidirectional synaptic plasticity (Fig. 1).

The results of one recently published study addressing this issue, in some way support this hypothesis. Suarez and colleagues (Suarez et al. 2016) showed that although LIDs lead to increased spines density selectively in the D2-expressing SPNs, the associated hyperexcitability is dependent on D1 receptors activation.

Conclusion

DA innervation in the striatum supports many important behavioral and cognitive functions such as movement control, procedural learning and working memory (De Leonibus et al. 2007; Yin et al. 2009; Cools et al. 2008). These behavioral functions require not only synaptic and structural plasticity but also metaplasticity (Giordano et al. 2018), which is a set of homeostatic mechanisms that confer to the synaptic matrix the necessary degree of flexibility to adjust to the dynamic changes in incoming inputs.

The findings we reviewed here show that DA denervation leads to a loss of synaptic and structural plasticity in corticostriatal pathway. DA restoration with l-DOPA has beneficial effects on behavior as long as its action can be modulated by spared pre-synaptic mechanisms of DA re-uptake and impulse-dependent release and by flexible dynamic regulation of direct and indirect pathways.